The future of support: A conversation with a customer support director

This interview is part of the SparrowDesk Spotlight series, where we sit down with the people rethinking how customer support actually works.

AI is changing customer support. But is the goal simply to make support faster?

For Karen, Director of Customer Support at Top Hat, the opportunity is much bigger. AI can help teams create bandwidth, experiment with new ways of working, improve the quality of customer conversations, and rethink what the support role looks like.

In this conversation, Karen shares her perspective on where AI fits into customer support, what customers actually expect from AI, why CSAT doesn't tell the whole story, and how support teams can prepare for what's next.

AI has been part of support for years. What's different now?

Karen's perspective:

AI itself isn't new to customer support.

Automation has existed for years in different forms, whether that's flagging and labeling emails or routing conversations. What's changed is the way we think about AI today, particularly with the rise of large language models.

When I think about AI in support, I encourage leaders to think about it in three different ways: doing things faster, doing things better, and doing things differently.

Speed and automation are important. They can help teams build bandwidth and create more space to work on other things.

But if we spend all our time focusing on speed and automation, we're missing opportunities to develop the support experience.

Doing things better and doing things differently are where you can take it to the next level.

Do customers actually care whether they're talking to AI or a human?

Karen's perspective:

I think customers care more about whether they're getting the resolution they're looking for.

If AI is able to unblock them or adequately provide the resolution they're looking for, I think they would be on board.

The question is also interesting because customers don't necessarily want to communicate with support in the first place.

Usually, when someone reaches out to support, there's already a reason for it. There might be a gap in the product, something that's confusing, or an issue they haven't been able to resolve themselves.

If a chatbot or self-service experience can give them the answer they need, they're perfectly happy that they didn't have to speak to a human.

But there will always be customers who prefer speaking to a person. Even before AI, customers had different preferences for chat, email, SMS, or phone.

So it's less about choosing between AI and humans and more about understanding how customers want to communicate and making sure the experience actually works for them.

Should support teams focus on deflecting as many conversations as possible?

Karen's perspective:

I don't think deflection should be the core problem support teams are trying to solve.

If a customer is already coming to support frustrated, our job isn't necessarily to deflect them. The core problem is figuring out how we can amplify the experience for that customer.

There was a flaw or shortcoming somewhere that led them to support. It could be product-related, service-related, or even connected to a previous interaction with the support team.

So the question becomes: how can we make sure their experience is better after that point of contact?

That's what I think support teams need to focus on.

How does AI change the role of a support agent?

Karen's perspective:

The expectations are changing.

If AI is handling some of the simpler conversations, support agents are increasingly dealing with more complex situations. That means there are higher expectations around technical ability, but also around skills such as de-escalation and objection handling.

Those skills aren't new. Support agents have always had them.

But when you add another layer of complexity to the work, agents may have to use those skills much more frequently.

We're also seeing new types of roles emerge within support. There are people whose work is focused on improving or monitoring AI models, looking at data, and assessing what can be done to improve the customer experience.

So it's not simply that the support representative role is changing.

The different types of roles within the support organization are changing too.

Is CSAT still enough to measure support quality?

Karen's perspective:

CSAT is useful, but it doesn't tell you everything.

One limitation is that it requires customers to actually provide their satisfaction score. Not every customer is going to take that extra step.

We also tend to see a lot of very happy or very unhappy responses, while the middle can get missed. NPS has a similar challenge.

Even metrics like first response time don't necessarily paint the whole picture. They tell you about speed, but not necessarily about the quality of the conversation.

What you really want to understand is the quality of the conversations your AI agents or human agents are having with customers.

Quality is difficult because it's subjective, but this is actually an area where AI could be helpful.

AI can analyze conversations at scale, provide sentiment, and help teams understand how those conversations went instead of relying entirely on customers to provide a score.

That's an opportunity to use AI to do support better, not just faster.

What other metrics should support leaders look at?

Karen's perspective:

One thing I'd encourage leaders to look at is case drivers.

Case drivers can be a shared metric between support and product and engineering because customers generally don't want to contact support unless there's a reason.

If we start seeing case drivers decline, that's a great indication that the product may be improving. From a service perspective, it could also mean there's less need for customers to reach out.

It's also important to remember that fewer conversations aren't automatically better.

If customers are having a good experience with your support team, they're actually more likely to communicate with you again. That's not necessarily a bad thing. It can mean they value and appreciate the support team.

There's also what we call shadow support, where customers use an LLM or AI search to find answers themselves.

That's why documentation has become increasingly important. If customers are going to use AI to find answers, you want to make sure the information they're finding is accurate and reflects what your company actually supports.

How should companies approach implementing AI in customer support?

Karen's perspective:

I would always encourage internal testing and implementation first, using smaller use cases and smaller groups.

Once you put a chatbot or any customer-facing AI experience out there, testing needs to continue after launch.

There will inevitably be drift. LLMs are updated over time, and because the outputs are non-deterministic, something that worked a certain way previously may not produce exactly the same outcome later.

That's why somebody needs to own and monitor the experience.

You also can't anticipate every possible use case before launch. Humans are unpredictable, and customers can use products in ways you never expected.

So you need to put things into practice, gather those use cases, learn from them, and continue improving.

That doesn't mean we shouldn't deploy AI to customers.

It means we need to understand what we're deploying, why we're deploying it, and how we're going to monitor it.

What are you most excited about for the future of customer support?

Karen's perspective:

For the first time in a really long time, our industry is changing, which is both scary but also incredibly exciting.

For well over a decade, not much changed with support. There were a few main ticketing systems, and many support teams operated from established playbooks.

Now we have more technology and tools available, and there are things we can explore that we may not have had the time or ability to focus on before.

For example, we've worked collaboratively within my team to build internal automations. One of them helps with logging customer bugs. Previously, that process could create a bottleneck before an issue even reached engineering.

Now we have internal tooling that makes the process much faster. That gives the person on my team more time to review the work and focus on things that require their attention.

That's what I'm excited about.

Taking the time to explore, experiment, and improve the customer experience.

What should support leaders keep in mind as AI evolves?

Karen's perspective:

AI is an opportunity.

I would encourage every leader to take the time to understand what AI can and cannot do, how they're using it, and why they're using it.

We also need to understand how it affects our roles and our teams. That doesn't necessarily mean the impact is negative. It means we need to understand what it means for us and for the company.

Once we have that understanding, we're able to be much better at our jobs.

Support is going to be around for a really long time. How the support team looks is what's changing.

And I think that's what makes this such an exciting time for the industry.